Population

People with atrial fibrillation or heart valve disease for whom long-term vitamin K antagonist therapy was required.

Setting

Self-INR monitoring supervised by primary or secondary care.

Interventions

The point-of-care devices considered in this assessment were:

• CoaguChek system
• INRatio2 PT/INR monitor
• ProTime Microcoagulation system.

Comparators

The comparator considered in this assessment was standard practice, which consisted of INR monitoring managed by health-care professionals. INR monitoring can be carried out in primary care, in secondary care or in a ‘shared provision’ setting:

• Primary care: INR monitoring can be carried out in primary care anticoagulant clinics using point-of-care tests or laboratory analysers. In the latter, blood samples are sent to a central laboratory based at a hospital (shared provision).
• Secondary care: INR monitoring can be carried out in hospital-based anticoagulant clinics using point-of-care tests or laboratory analysers.

Outcomes

The following outcomes were considered.

Number and type of studies excluded

Appendix 5 lists the number of studies excluded after full-text assessment and the reasons for their exclusion.

Selection bias

Of the 26 included trials, only seven trials reported adequate details on both generation of the randomisation sequence generation and concealment of allocation.^{55,57,61–65} In 11 trials, the randomisation process proved to be adequate but no information was provided on the way in which participants were allocated to the study interventions.^{58,60,66–74} One trial⁷⁵ reported adequate details about the generation of the random sequence but failed to conceal the allocation of participants to study interventions. In contrast, another trial⁷⁶ reported adequate information on allocation concealment but failed to provide details on the randomisation process. In six trials, both the randomisation process and the allocation concealment were judged as ‘unclear’ due to the lack of adequate information.^{45,56,59,77–79}

Attrition bias

Seventeen trials were judged to be at low risk of attrition bias. Six of them had limited missing data with similar reasons for discontinuation across intervention groups.^{59,69,72,74,76,79} Seven trials relied on an intention-to-treat approach and all dropouts were fully accounted for in the statistical analyses,^{55,61,63–65,71,75} while the other four reported no missing data.^58,60,62,78 Eight of the 26 included trials were at high risk of attrition bias, with more than 5% dropout rate and with missing data not appropriately tackled.^{45,56,57,66–68,73,77} In the Early Self-Controlled Anticoagulation Trial,⁷⁰ the problem of incomplete outcome data was addressed for the first 600 participants but not for all included participants.

Performance and detection bias

Owing to the nature of the interventions being studied (use of point-of-care devices), blinding of participants or personnel was not feasible. Seven trials blinded the outcomes assessor (statistician or clinical outcome assessor).^{55,58,61–63,72,77} In six trials, neither the participants nor the personnel involved in delivering the interventions were blinded.^{60,65,66,71,74,75} One trial⁷⁰ was described as ‘double blinded’ but no further information was given. Another trial⁶⁸ reported that one of the two standard care groups studied (the untrained routine group) was blinded. In addition, this trial revealed that the nurses involved in transferring data on dosing as well as the dosing physicians were blinded. The remaining of the included trials did not provide information on blinding.

Reporting bias

With the exception of three trials,^58,76,79 the outcomes reported in the trials were prespecified in the analysis section and reporting bias was not obvious in the published papers.

Other sources of biases

No other sources of biases were obvious in the included trials.

Study details

The majority of included trials were conducted in Europe: six trials were conducted in Germany,^{57,59,60,70,72,78} six in the UK,^{45,56,64,67,69,73} three in Denmark,^58,66,75 three in the Netherlands,^68,76,79 one in Ireland,⁶² one in Austria,⁶³ one in France⁷⁷ and one in Spain.⁶¹ Three trials were conducted in Canada^55,65,74 and one in the USA.⁷¹ Of the 26 included trials, seven were multicentred,^{60,63,64,67,68,71,72} while the remaining 19 were conducted in a single centre.

The length of follow-up ranged from 14 weeks⁵⁷ to more than 4 years.^63,71 Nine trials reported follow-ups ≥ 12 months.^{55,61,63,64,70,71,73,78,79} One trial, which was originally supposed to run for 2 years, was discontinued prematurely due to the small number of recruited participants.⁶⁰

Nine of the included trials were funded independently by professional organisations or national/governmental agencies^{55,57–59,64–66,69,75} while 13 trials were fully or partly funded by industry. In the case of the remaining four trials, the source of funding was not reported.^70,76,78,79

Participants

Most of the included trials (15 out of 26) included participants with mixed indications of which atrial fibrillation, artificial heart valves (AHVs) and venous thromboembolism were the most common clinical indications,^{45,55–57,61–63,65,66,68,71,72,74–76} while six trials enrolled exclusively participants with AHVs^{59,70,73,77–79} and two trials limited inclusion to participants with atrial fibrillation.^60,69 Seven trials provided information on risk factors, comorbidity, and/or previous bleeding and thromboembolic events but did not report significant baseline differences between participants in self-monitoring and those in standard care (see Appendix 7, Table 32).^{61,63,64,71,72,75,77}

The mean sample size among the included trials was 337 participants (range 1639–292,271 participants). Fifteen trials performed a power analysis and a sample size calculation,^{58,60–66,68,69,71,72,74–76} two trials, with very small sample sizes, did not power their studies^55,57 and the remaining trials did not provide information on how the sample size was determined.^{45,56,59,67,70,73,77–79} The age of adult participants ranged from 16 to 91 years.⁶² The only trial which assessed children reported a median age of 10 years.⁵⁵

Warfarin was the choice of vitamin K antagonist therapy in half of the included trials.^{45,55,56,58,62,64–67,69,71,73,74,78} In seven trials, participants were taking phenprocoumon and/or acenocoumarol and/or fluindione^{57,61,63,68,72,76,77} and, in one trial, participants received either warfarin or phenprocoumon.⁷⁵ In the remaining four trials, the type of vitamin K antagonist therapy was not reported.^59,60,70,79 In nearly half of the included trials (12 out of 26), participants had been on OAT for at least 3 months before randomisation.^{45,55,56,61,64,66–69,74–76} Three trials included vitamin K antagonist-naive participants for whom long-term OAT was recently indicated but who had not been on anticoagulation therapy before.^58,63,70 In the largest trial, The Home International Normalised Ratio Study (THINRS),⁷¹ randomisation was stratified according to the duration of anticoagulation but no significant differences were found between participants who had started anticoagulation therapy within the previous 3 months and those who had received anticoagulation therapy for > 3 months. In the two remaining trials, the included participants received OAT for < 3 months (1–2 months) before randomisation.^62,65

Point-of-care tests used for international normalised ratio measurement

CoaguChek system for INR monitoring was used in 22 of the 26 included trials. Nine trials used the S model,^{45,56,58,61,63,64,67,75,78} four used the XS model,^55,62,66,74 one⁷⁰ used the CoaguChek Plus model and two trials used the first model of the CoaguChek series, which was simply referred to as ‘CoaguChek’.^68,72 In six trials, it was unclear whether the CoaguChek device was the first model or its later versions.^{59,60,69,73,76,79} Either the INRatio or the CoaguChek S was used for INR measurement in two trials (but results were not separated according to the type of the point-of-care device),^57,77 and the ProTime system was used in other two trials.^65,71 In all six trials based in the UK, CoaguChek system (either CoaguChek or version S) was used for the INR measurement.

In 11 trials, in order to assure accuracy of the point-of-care devices being used, INR results measured directly by participants were compared with those measured in a laboratory.^{45,55–58,64–67,69,77}

Eight trials’ investigators who did not specify the model of the CoaguChek device (S or XS) used for INR measurement were contacted for further details.^{60,68,69,72,73,76,77,79} Five of the them provided further information on the model of the CoaguChek point-of-care device.^{58,68,72,77,79}

Standard anticoagulant management

The type of standard care varied across trials. In 13 trials, INR was measured by professionals in anticoagulant or hospital outpatient clinics,^{45,57,58,61,62,64,66,68,69,71,74,76,79} by a physician or a GP in a primary care setting in six trials,^{59,60,65,67,70,78} and either by a physician/GP in a primary care setting or by professionals in anticoagulant/outpatient hospital clinics in five trials.^{63,72,73,75,77} In two trials, comparing self-testing with self-management,^55,56 self-testing within anticoagulant clinics was considered as standard care. In the majority of the included trials (17 out of 26), the anticoagulant clinic was led by a clinician (general or specialist),^{58–61,63,65,66,68–73,75,77–79} by a nurse in five trials, (three conducted in the UK,^45,48,64 one in Canada⁵⁵ and one in Germany⁵⁷) and by a pharmacist in two trials, conducted in Canada⁷⁴ and in Ireland.⁶²

International normalised ratio measurement was carried out in a laboratory in all but two trials, where CoaguChek S⁶⁷ or another coagulometer⁷⁵ was used instead.

Self-monitoring

The majority of the included trials (17 out of 26) compared self-management (participants performed the test and adjusted the dose of anticoagulation therapy themselves) with standard care,^{57–61,63–65,67,70,72–76,78,79} six assessed self-testing (participants performed the test themselves with the results managed by health-care professionals)^{45,62,66,69,71,77} and one evaluated both self-testing and self-management versus either trained or untrained routine care (four arms).⁶⁸ It is worth noting that for the subgroup meta-analysis according to type of OAT management, this four-arm trial contributed to two studies: one on self-testing and one on self-management. The two standard care groups (trained and untrained routine care) were initially combined to produce an overall control group and subsequently subdivided into two groups for the purpose of the subgroup analysis, which was undertaken to assess the effects of self-testing versus self-management.

The remaining two trials compared self-testing with self-management (without standard care as a comparator).^55,56 One of these two trials enrolled exclusively a population of children⁵⁵ while the other provided a non-randomised comparison of participants in self-testing and self-management with those receiving standard care for a period of 6 months before study enrolment.⁵⁶ We deemed these two trials suitable for inclusion as they provide relevant outcomes for participants in both self-testing and self-management.

In 19 out of 26 included trials, participants received training and education in order to perform self-testing and self-management (see Appendix 7, Table 33).^{45,55,56,61,62,64–67,69,72–77,79–81} In most of these trials (11 out of 19), the training was provided in group sessions which lasted for around 1–2 hours^{55,61,62,68,69,72,76,77,81} up to a maximum of 3 hours.^73,74 The training was usually administered by a single member of staff, either a nurse, a practitioner/physician^{45,56,61,64,69} or a pharmacist.⁷⁴ In a few trials, the training was provided by a team of professionals, such as a specialist physician together with paramedical personnel,⁸⁰ a research pharmacist coupled with an haematologist⁶² or a physician assisted by a nurse.^63,72 In five trials, the personnel responsible for delivering the training was reported to be trained specifically on self-testing and self-management.^{56,61,63,64,72}

Overview

This section provides evidence from 26 included trials on the clinical effectiveness of self-monitoring using CoaguChek system, INRatio2 PT/INR monitor and ProTime Microcoagulation system, compared with standard practice (see Figures 3–14; Tables 4–6; Appendix 8). For clarity, the results are reported under the broad headings of ‘clinical outcomes’, ‘intermediate outcomes’ and ‘patient-reported outcomes’. The summary effects of relevant clinical outcomes such as bleeding events, thromboembolic events and mortality have been described separately within the ‘clinical outcomes’ section. Tables 4 and 5 show the main findings of the five trials conducted in the UK and of the four trials using CoaguChek XS. The results of the sensitivity analyses for each point-of-care test are displayed in Table 6.

FIGURE 3

Forest plot of comparison: any bleeding events. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus; PSM, patient self-management; PST, patient self-testing.

FIGURE 14

Forest plot of comparison: mortality – type of standard care. AC, anticoagulation; C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

TABLE 4

Main findings of the five UK trials

TABLE 6

Results of sensitivity analyses according to the type of point-of-care device (CoaguChek/INRatio/ProTime)

TABLE 5

Main findings of the four trials using CoaguChek XS

Clinical outcomes

Bleeding

Twenty-one trials reported a total of 1472 major and minor bleeding events involving 8394 participants.^59–79 Two trials reported that there were no bleeding events and hence did not contribute to the overall effect size in the related meta-analysis.^66,75 Twenty-one trials reported 476 major bleeding events in a total of 8202 participants,^{59–65,67–74,77–79} while 13 trials reported 994 estimable minor bleeding events in a total of 5425 participants.^{61,63,64,67,69,71–74,76,77} No statistically significant differences were observed between self-monitoring participants (self-testing and self-management) and those in standard care for any bleeding events (RR 0.95, 95% CI 0.74 to 1.21; p = 0.66) (Figure 3), major bleeding events (RR 1.02, 95% CI 0.86 to 1.22; p = 0.80) (Figure 4) and minor bleeding events (RR 0.94, 95% CI 0.65 to 1.34; p = 0.73) (Figure 5). The results were not affected by the removal of the UK-based trials (see Appendix 8) or by the removal of the trials assessing ProTime and/or INRatio (see Table 6 and Appendix 8). Similarly, sensitivity analyses restricted to CoaguChek XS trials demonstrated no differences from the all-trials results (see Table 6 and Appendix 8). A sensitivity analysis restricted to trials at low risk of bias slightly changed the estimate of effect but did not significantly impact on the findings (RR 0.59, 95% CI 0.27 to 1.30; p = 0.19) (see Appendix 8).

FIGURE 4

Forest plot of comparison: major bleeding events. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus; PSM, patient self-management; PST, patient self-testing.

FIGURE 5

Forest plot of comparison: minor bleeding events. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

The subgroup analysis by type of anticoagulant management therapy did not show any difference between self-management and standard care for any bleeding events (RR 0.94, 95% CI 0.68 to 1.30; p = 0.69) but revealed a significant higher risk in self-testing participants than in those receiving standard care (RR 1.15, 95% CI 1.03 to 1.28; p = 0.02) (see Figure 3). When trials assessing ProTime and INRatio were removed from the analysis, a non-significant trend was observed in favour of self-testing (0.58, 95% CI 0.22 to 1.47; p = 0.25) (see Table 6 and Appendix 8). No significant differences in the risk of major bleeding were observed between self-management (RR 1.09, 95% CI 0.81 to 1.46; p = 0.58) and self-testing (RR 0.99, 95% CI 0.80 to 1.23) versus standard care (see Figure 4). When only minor bleeding events were assessed (see Figure 5), a significant increased risk was observed in self-testing participants (23%), compared with those in standard care (RR 1.23, 95% CI 1.06 to 1.42; p = 0.005), but not in those who were self-managed (RR 0.84, 95% CI 0.53 to 1.35; p = 0.47). Two trials enrolled participants with atrial fibrillation, six trials enrolled participants with AHVs and 13 trials enrolled participants with mixed indication. No statistically significant subgroup differences were found for bleeding events according to the type of clinical indication (Figure 6). Similarly, for bleeding events, no significant differences were detected when trials were grouped according to the type of control care (anticoagulant clinic care RR 0.84, 95% CI 0.50 to 1.42; p = 0.52; GP/physician RR 1.09, 95% CI 0.79 to 1.50; p = 0.60; mixed care RR 0.94, 95% CI 0.79 to 1.13; p = 0.54) (Figure 7).

FIGURE 6

Forest plot of comparison: any bleeding events – clinical indication. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

FIGURE 7

Forest plot of comparison: any bleeding events – type of standard care. AC, anticoagulant; C, CoaguChek; CS, CoaguCchek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

Thromboembolic events

Twenty-one trials reported 351 major and minor thromboembolic events in a total of 8394 participants.^59–79 Six of these trials did not contribute to the overall estimate of effect as they reported ‘zero’ events in both groups.^{66–69,74,75} Self-monitoring (self-testing and self-management) showed a statistically significant reduction in the risk of thromboembolic events by 42% (RR 0.58, 95% CI 0.40 to 0.84; p = 0.004), compared with standard care (Figure 8). The risk reduction further increased to 48% when only major thromboembolic events were considered (RR 0.52, 95% CI 0.34 to 0.80; p = 0.003) (Figure 9). The risk of thromboembolic events significantly decreased when the analyses were restricted to non-UK trials (RR 0.50, 95% CI 0.32 to 0.76; p = 0.001); to CoaguChek trials (RR 0.52, 95% CI 0.38 to 0.71; p < 0.0001); and to trials at low risk of bias (RR 0.38, 95% CI 0.16 to 0.92; p = 0.03) (see Appendix 8).

FIGURE 8

Forest plot of comparison: thromboembolic events. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus; PSM, patient self-management; PST, patient self-testing.

FIGURE 9

Forest plot of comparison: major thromboembolic events. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus; PSM, patient self-management.

Self-management compared with standard care halved the risk of thromboembolic events (RR 0.51, 95% CI 0.37 to 0.69; p < 0.0001). In contrast, for self-testing participants^59,62,71,77 no significant risk reduction was observed compared with those in standard care (RR 0.99, 95% CI 0.75 to 1.31; p = 0.95) (see Figure 8). The subgroup difference between self-management and self-testing was statistically significant (p = 0.002). When trials assessing the ProTime system were removed from the analysis, the risk reduction increased from 1% to 45% but the summary estimate of effect was not statistically different from the all-trials summary estimate (RR 0.55, 95% CI 0.13 to 2.31; p = 0.41) (see Table 6 and Appendix 8). Self-monitoring participants with AHVs showed a significant reduction in the number of thromboembolic events, compared with those in standard care (RR 0.56, 95% CI 0.38 to 0.82; p = 0.003). Among participants with mixed clinical indication (atrial fibrillation, AHVs or other conditions), the effect was larger but not statistically significant than that observed in participants receiving standard care (RR 0.57, 95% CI 0.30 to 1.09; p = 0.09) (Figure 10). The risk of thromboembolic events reduced in self-monitoring participants by 55% when routine anticoagulation control was managed by a GP or physician (RR 0.45, 95% CI 0.29 to 0.68; p = 0.0002). In contrast, even though fewer thromboembolic events were observed in participants who self-monitored their therapy than in those managed in specialised anticoagulation clinics (RR 0.65, 95% CI 0.30 to 1.42; p = 0.28) or those in mixed provision managed by either a physician/GP or a specialist (RR 0.66, 95% CI 0.31 to 1.38; p = 0.27), no significant subgroup differences were detected (Figure 11).

FIGURE 10

Forest plot of comparison: any thromboembolic events – clinical indication. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

FIGURE 11

Forest plot of comparison: any thromboembolic events – type of standard care. AC, anticoagulation; C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

Mortality

Thirteen trials reported 422 deaths due to all-cause mortality in a total of 6537 participants.^{61,63,64,66,67,70–75,77,79} Two trials with zero fatal cases did not contribute to the overall estimate of effect.^66,74 One trial of 1200 participants⁷⁰ reported overall mortality data without separating the results for participants self-managed and for those receiving standard care. We contacted the corresponding author of this trial for further information but we did not receive any reply. Therefore, for mortality data, for this particular trial, we relied on the estimates published in the previous meta-analysis by Garcia-Alamino and colleagues³¹ and in the HTA by Connock and colleagues,²¹ which were based on individual patients’ data.

The risk reduction for all-cause mortality was not statistically significant different between self-monitoring (self-testing and self-management) and standard care (RR 0.83, 95% CI 0.63 to 1.10; p = 0.20) (Figure 12). The results were not affected by the removal of the UK-based trials or by the removal of trials at high or unclear risk of bias. When the analysis was restricted to trials that used the CoaguChek system, the summary estimate for self-monitoring was not different from the all-trials estimate (RR 0.68, 95% CI 0.46 to 1.01, p = 0.06) (see Table 6 and Appendix 8). Two trials reported six deaths out of a total of 932 participants, related to vitamin K antagonist therapy.^61,63

FIGURE 12

Forest plot of comparison: mortality. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

Risk of death reduced by 32% through self-management (RR 0.68, 95% CI 0.46 to 1.01; p = 0.06) but not through self-testing (RR 0.97, 95% CI 0.78 to 1.19; p = 0.74), even though the test for subgroup differences was not statistically significant (p = 0.13) (see Figure 12). Self-monitoring halved the risk of mortality in participants with AHVs (RR 0.54, 95% CI 0.32 to 0.92; p = 0.02) but not in those with mixed clinical indication for AOT (RR 0.95, 95% CI 0.78 to 1.16; p = 0.61) (Figure 13). The subgroup difference between participants with AHVs and those with mixed indication with regard to the number of deaths was statistically significant (p = 0.05). No data were available from trials that enrolled participants with atrial fibrillation. Significantly fewer deaths were recorded among participants who self-monitored their therapy than among those who were routinely managed by their GP/physician (RR 0.52, 95% CI 0.30 to 0.90; p = 0.02) (Figure 14).

FIGURE 13

Forest plot of comparison: mortality – clinical indication. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’; CPlus, CoaguChek Plus.

Anticoagulation control: target range

Anticoagulation control can be measured as the time that INR is in the therapeutic range or as INR values in therapeutic range. Data on INR TTR were available from 18 trials.^{55,56,58,60–69,71,73–75,77} However, there was variation in the measures used for reporting TTR. Seven trials comparing self-monitoring with standard care reported TTR as mean percentage;^{61,64,65,69,71,74,77} three as median percentage,^58,62,75 five as overall percentage^{63,66–68,73} and one as cumulative number of days.⁶⁰ The two remaining trials, which compared patient self-management with patient self-testing, reported the TTR as mean percentage time (one trial)⁵⁵ and overall percentage time (the other trial).⁵⁶ It proved impossible to convert median values into mean values due to the lack of information on the maximum or minimum value required by the conversion formula. Therefore, we were unable to pool the TTR results from the 18 trials which provided this information. The results of these trials are shown in Table 7.

TABLE 7

International normalised ratio results (TTR and INR values in target range)

Time in therapeutic range ranged from 52%⁵⁸ to 80%^66,74 for self-monitoring and from 55%⁵⁸ to 77%⁶⁷ for standard care. In all but three trials,^58,61,67 TTR was higher in self-monitoring participants than in those receiving standard care and, in five of these trials, the difference between intervention groups was statistically significant.^{62,66,71,73,77} Three of the UK-based trials reported no significant differences between self-monitoring and standard care.^64,67,69 Pooling of results was possible for 10 trials that provided suitable data.^{61,64–69,71,74,77} No statistically significant differences were found between self-management and standard care (RR 0.47, 95% CI –1.40 to 2.34; p = 0.62). A modest but significantly higher proportion of TTR was, however, found for participants assigned to self-testing than for those in control care (WMD 4.44, 95% CI 1.71 to 7.18; p = 0.001) (Figure 15). It is worth noting that the overall estimate of effect was dominated by the largest included trial on self-testing, THINRS.⁷¹ In two trials, one using CoaguChek XS⁶⁶ and the other using ProTime,⁷¹ the WMD between self-testing and standard care for TTR was significantly higher, indicating better anticoagulation control among self-testing participants.

FIGURE 15

Forest plot of comparison: TTR. C, CoaguChek; CS, CoaguChek ‘S’; CXS, CoaguChek ‘XS’.

The INR values in therapeutic range were reported in 12 trials.^{59–61,64,66–68,70,72,76,78,79} There was great variation between trials in the measures used to assess INR values in therapeutic range and, therefore, the pooling of data across trials proved unfeasible. In eight trials which reported the proportion of INR measurements in the therapeutic range,^{59,60,66–68,70,76,78} the values ranged from 43.2%⁵⁹ to 80.8%⁶⁶ for self-monitoring and from 22.3%⁵⁹ to 72%⁶⁷ for standard care. In four trials that reported the proportion of participants in the therapeutic range instead,^61,64,72,79 the values ranged from 53%⁷² to 72.9%⁷⁹ for self-monitoring and from 43.2%⁷⁹ to 72%⁶⁴ for standard care. With the exception of two UK-based trials,^64,67 all trials reported higher proportion of INR measurements or larger proportions of participants in therapeutic range for self-monitoring than for standard care. Significant differences between interventions were detected in six of these trials.^{60,61,66,70,78,79} The INR values in therapeutic range are summarised in Table 7.

Among participants with AHVs, self-monitoring resulted in a significantly higher INR TTR^73,77 or INR values in therapeutic range^59,70,78,79 than standard care. In two trials that included participants with atrial fibrillation,^60,69 no TTR differences were found between self-monitoring and standard care.

Test failure rate

Only one trial⁴⁵ reported one instrument defect and one test strip problem in the self-testing group. No other failures were mentioned in the remaining included trials.

Time to receive test result

One trial⁷⁴ reported the time for each INR monitoring (i.e. time from INR measurement to test results) and the total time spent for anticoagulant management during the 4-month follow-up period. The time spent for each INR monitoring by self-managed participants was significantly lower (mean 5.3 minutes, SD 2.6 minutes) than the time spent by participants receiving standard care (mean 158 minutes, SD 67.8 minutes; p < 0.001). During the 4-month follow-up, the total time spent for anticoagulation monitoring by participants in standard care was significantly higher (mean 614.9 minutes, SD 308.8 minutes) than the total time spent by participants who self-managed their therapy (mean 99.6 minutes, SD 46.1 minutes; p < 0.0001).

Patient compliance with testing

Gardiner and colleagues⁴⁵ reported > 98% compliance with self-testing and stated that participants were conscientious in performing and recording their weekly tests. Of those who did not comply with self-testing, two had difficulties performing the test or experienced disruption due to hospitalisation and one lost the CoaguChek meter. In the trial by Khan and colleagues,⁶⁹ 75% (30 out of 40) of participants did not report any problems with the use of the device and expressed willingness to continue with self-monitoring. On the other hand, participants who did not comply (25%) with the testing procedure reported difficulties with the technique or problems placing the fingertip blood drop on the right position on the test strip. This resulted in the need to use multiple strips to achieve a single reading.

Frequency of testing

Even though the frequency of self-testing was preplanned in 18 of the included trials,^{45,56,59,62–69,71–77} only 10 trials eventually reported it.^{59,62–68,71,76} The frequency of self-testing ranged on average (mean) from every 4.6 days⁶² to every 12.4 days⁶⁴ (Table 8).

TABLE 8

Frequency of self-testing and adherence to self-monitoring

Frequency of visits to primary or secondary care clinics

Frequency of visits to clinics was reported by 12 trials. Three trials reported three visits in approximately 6 months;^62,69,72 five trials reported four visits per year;^{59,64,67,71,79} three trials reported two visits per year;^63,66,78 and the remaining trial⁷³ reported that there were no routine clinic visits during the study period (see Table 8).

Adherence to the self-monitoring

Generally, adherence to the self-monitoring was reported to be high in the included trials. In 13 trials, > 90% of the participants completed self-monitoring,^{58,59,61–63,66,69,71,72,74–76,79} while in another nine trials the dropout rates ranged from 12% to 25% (see Table 8).^{45,55–57,64,65,67,73,77} Few trials also provided information on the number of people who were excluded during the recruitment phase because they were incapable of performing self-monitoring.^{45,56,61,67,71,73,77,80} It is worth noting that up to 50% of the participants were deemed unsuitable for inclusion. Main reasons for exclusion were old age, anxiety, lack of confidence, inadequate cognitive abilities and poor dexterity.

People’s anxiety associated with waiting time for results and with not knowing their current coagulation status and related risk

The trial by Bauman and colleagues,⁵⁵ which compared self-management with self-testing in children, reported that one parent (single parent of a 16-year-old male child) did not favour self-management because of the increased anxiety related to INR measurements.

Preference and acceptability of the tests

Four trials^45,55,57,62 conducted a questionnaire survey to assess acceptability to participants of self-testing and self-management using point-of-care devices (Table 9). These trials reported high rates of acceptance for both self-management and self-testing (77% to 98%).^45,55,57,62 Two trials^45,62 reported that 77% to 98% of participants favoured self-testing with CoaguChek S over standard care.

TABLE 9

Acceptability of the tests

Another crossover trial⁵⁷ reported that 93% of participants rated their satisfaction with regard to self-monitoring (using either INRatio or CoaguChek S) as high or good. When asked about the overall relative satisfaction with the device, 43% of participants favoured INRatio, 36% favoured CoaguChek S, and 21% liked both devices equally. The trial by Bauman and colleagues,⁵⁵ which assessed self-management over self-testing (usual care in this trial) in children, reported that the majority of participants (13 out of 14 participating families: 92%) opted for the use of CoaguChek XS device.

Health-related quality of life

Nine trials^{55,67–69,71,72,74,76,79} reported on health-related quality-of-life outcomes using one of the following measures:

1. Sawicki’s tool:^68,72,74,76 a structured questionnaire containing 32 questions developed and validated by Sawicki and colleagues.⁷² The questionnaire covered five treatment-related aspects: ‘general treatment satisfaction’, ‘self-efficacy’, ‘daily hassles’, ‘distress’ and ‘strained social network’. The questions were derived from the sentences formulated by the participants receiving anticoagulation describing their feelings with regard to their treatment. Each item is graded on a scale ranging from a minimum of 1 (total disagreement) to a maximum of 6 (total agreement) as self-perceived by participants. Higher scores for self-efficacy and general treatment satisfaction and lower scores for daily hassles, psychological distress and strained social network are indicative of better quality of life.
2. Short Form questionnaire-36 items (SF-36) [UK Short Form Health Survey (UKSF-36), SF-36v2]:^69,79 SF-36 is a validated tool containing 36 items for the assessment of the health status and quality of life. SF-36 covered physical functioning, physical role limitation, bodily pain, general health perceptions, vitality, social functioning, emotional role limitation and mental health. UK SF-36 and the SF-36v2 questionnaire have been reported here.
3. Euroqol scores [European Quality of Life-5 Dimensions (EQ-5D)]:⁶⁹ EQ-5D is a validated tool for assessing health status and quality of life.
4. Lancaster’s instrument:^67,69 Lancaster’s instrument is designed to measure health beliefs specific to the use of warfarin in anticoagulant treatment.
5. Duke Anticoagulation Satisfaction Scale:⁷¹ Duke Anticoagulation Satisfaction Scale measures patient satisfaction with anticoagulation. Scores on this scale range from 25 to 225, with lower scores indicating higher satisfaction.
6. Health Utilities Index Mark 3 (HUI Mark 3):⁷¹ HUI Mark 3 is a tool to measure quality of life. Scores for the HUI Mark 3 range from −0.36 to 1.00, with a negative score indicating a state worse than being dead and a score of 1.00 indicating perfect health.
7. Schedule for Evaluation of Individual Quality of Life (SEIQoL) tool:⁶⁷ SEIQoL is a semistructured interview for the assessment of quality of life.
8. KIDCLOT-PAC-QL: KIDCLOT-PAC Parent-proxy QL^© (parents’ quality of life and their assessment of the child’s quality of life) and KIDCLOT-PAC Child-teen QL^©.⁵⁵

Four trials reported quality of life using Sawicki’s questionnaire (Table 10).^68,72,74,76 Sawicki and colleagues⁷² and Verret and colleagues⁷⁴ reported improvements in treatment satisfaction and self-efficacy, and reduced level of distress and daily hassles in both the self-management and the standard care groups, but the improvements were significantly greater among participants self-managed. Similarly, Cromheecke and colleagues⁷⁶ showed significant improvements in treatment satisfaction and self-efficacy, and significant reductions in level of distress and daily hassles for self-management participants, compared with those in standard care. Gadisseur and colleagues⁶⁸ showed increased treatment satisfaction and self-efficacy, and reduced level of distress and daily hassles among self-monitoring participants (self-testing or self-management). On the other hand, they found an increased level of distress among participants who received education but did not directly monitor their anticoagulation therapy.

TABLE 10

Quality of life measured using Sawicki’s tool

Two UK-based trials did not find significant differences in quality-of-life outcomes between self-management and self-testing participants, compared with those receiving standard care (Table 11).^67,69 Khan and colleagues⁶⁹ reported quality-of-life data using the UK SF-36, the Euroqol scores and Lancaster’s instrument. No significant differences were observed between self-testing participants and those who received education but did not test themselves, for both the UK SF-36 parameters and the Euroqol scores. Emotional function was the only parameter that showed a significant change at 24 weeks compared with baseline (p = 0.04). Fitzmaurice and colleagues⁶⁷ assessed participants’ attitude towards self-management and quality-of-life outcomes through a semistructured interview given to a random sample of 16 participants (eight in self-management and eight in standard care). Assessed themes were adapted from Lancaster’s instrument, the SEIQoL tool and a series of focus groups. Five common themes emerged from the interviews conducted on participants in self-management: knowledge and management of condition, and self-empowerment, increased anxiety and obsession with health, self-efficacy, relationship with health professionals, and societal and economic cost. The trial investigators did not find any significant difference in quality of life between participants self-managed and those in standard care. Soliman Hamad and colleagues⁷⁹ measured quality of life in participants with AHVs in the Netherlands by means of the SF-36v2. Significant improvements in quality-of-life scores were observed in participants who self-managed their therapy, compared with those in standard care, with regard to the physical component summary only (see Table 11).

TABLE 11

Quality of life measured using SF-36

Matchar and colleagues⁷¹ measured quality of life by means of the HUI Mark 3. They reported significant gain in health utilities at the 2-year follow up among self-testing participants who used ProTime, compared with those managed in high-quality anticoagulant clinics (p < 0.001). The same investigators⁷¹ also measured anticoagulant satisfaction using the Duke Anticoagulation Satisfaction Scale. They found that the degree of satisfaction was higher in self-testing participants than in those in standard care (p = 0.002).

Bauman and colleagues⁵⁵ assessed self-management versus self-testing in children and provided quality-of-life data using the KIDCLOT-PAC Parent-proxy QL^© (parents’ quality of life and their assessment of the child’s quality of life) and the KIDCLOT-PAC Child-teen QL^©.

Both tools were completed pre and post intervention to assess potential changes in quality-of-life outcomes related to warfarin use. The five common themes identified from the open-ended questionnaires and the semistructured interviews were awareness, communication, relationship between parent and child, flexibility and anxiety. No significant changes in ‘tasks’ related to warfarin use were found between intervention groups.